fix: raise ValueError instead of UnboundLocalError in _get_indexed_match on zero matches

[chuenchen309]

Problem

_get_indexed_match (unstructured/cleaners/extract.py) does for i, result in enumerate(re.finditer(pattern, text)), then references i in the "not found" branch's error message. When the pattern matches zero times, the loop body never runs, so i is never bound — the intended ValueError("Result with index ... was not found...") instead raises UnboundLocalError: cannot access local variable 'i', breaking extract_text_before()'s and extract_text_after()'s documented "pattern not found" contract.

(The existing test_get_indexed_match_raises_value_error_with_index_too_high test doesn't catch this — it uses a pattern that does match, just fewer times than the requested index, so i was always bound there.)

Fix

Initialize i = -1 before the loop, so the existing ValueError branch can safely report a sensible "largest index" (-1, meaning none) even with zero matches.

Testing

• Added test_get_indexed_match_raises_value_error_with_zero_matches (a pattern with zero matches in the text) alongside the existing index-too-high test.
• Confirmed red→green: reverting only extract.py reproduces UnboundLocalError: cannot access local variable 'i'; reapplying passes.
• Full test_unstructured/cleaners/ (140 tests) — all pass.
• ruff check — clean.
• xref: no open PR touches unstructured/cleaners/extract.py.

Note: this bumps CHANGELOG.md/__version__.py to 0.25.1, matching the same version already proposed by two of my other open PRs (#4397, #4398) and one already-merged... actually still-open (#4399) against this same base — all four independently propose the same next-patch version bump since none have merged yet. Whichever merges first should "win" the bump; the rest will need a trivial rebase, which I'm happy to do once the ordering is known.

AI-Generated disclosure

Found via an AI-assisted code review pass (Claude Code) over unstructured/cleaners/. I personally reproduced the UnboundLocalError, verified the fix, and ran the relevant test suite plus lint before submitting.

Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com

[cubic-dev-ai]

No issues found across 4 files

<sub>Shadow auto-approve: would auto-approve. Fixes a clear bug where zero matches raised UnboundLocalError instead of the intended ValueError; the change is a one-line initialization plus a regression test, bounded and clearly beneficial.

Re-trigger cubic</sub>